The first newspaper stories showed up in the summer of 1997, buried on the inside pages: an influenza virus designated H5N1, known to be fatal to chickens but never before seen in humans, had killed a 3-year-old child in Hong Kong. BOY'S DEATH STUMPS EXPERTS, read the small headline in The Atlanta Journal-Constitution. Then, in December, a 2-year-old in a different part of Hong Kong fell ill from the same virus, but recovered. "To me," remarked Robert G. Webster, a leading virologist, "the startling thing about the second case is that there is a second case." More cases appeared--18 in all, of which six were fatal--and Hong Kong prepared to slaughter every chicken within its borders. The danger passed, then flared up in 2001 and 2002, when chickens in the Hong Kong markets began dying again, although this time there were no known human victims. Millions of fowl were killed in an effort to eradicate the virus, but it cropped up again in 2003, this time in seven countries, including Thailand and Vietnam. The toll this year stands at 31 human deaths out of 43 cases, a fatality rate of 75 percent. Only a few diseases, such as untreated AIDS, have been as lethal. (SARS, an unrelated respiratory disease that caused so much panic last year, killed 774 out of about 8,100 victims.) So far, virtually all the victims of H5N1 appear to have caught it directly from contact with fowl. Should the virus somehow acquire the ability to spread from person to person, the result could be a global catastrophe. Says Dr. John Treanor, an infectious-disease expert at the University of Rochester, "I don't think we completely understand why it hasn't happened already."

This year's vaccine shortage may cost lives (and the jobs of those responsible), but it falls under the heading of administrative nightmares, not medical ones. Infectious-disease experts have a different scenario to keep them awake: a novel influenza virus to which people have never been exposed getting loose in a crowded world crisscrossed by jet planes that bring Tokyo, Nairobi and Moscow within, figuratively, spitting distance of one another. Three times in the last century--in 1918, 1957 and 1968--such a virus emerged, and the result each time was a "pandemic," an epidemic that's unusually widespread. The worst by far was in 1918, when at least 20 million people died. It was called the "Spanish flu," but a new book by John M. Barry, "The Great Influenza," argues that it actually emerged in rural Kansas, a place where, as in parts of present-day Asia, people lived in close proximity to domestic fowl and pigs.

The 1918 flu--variant H1N1--spread with terrifying speed; in six days at a single Army base, Barry writes, the hospital census went from 610 to more than 4,000. It killed with devastating swiftness: pedestrians literally collapsed in the street; people woke up healthy and were dead by nightfall. It attacked multiple organs in the body, but always the respiratory system first, laying waste to the defenses by which the body keeps pathogens out of the lungs. Most victims succumbed to a secondary infection of bacterial pneumonia, for which there was no treatment in 1918. But in other cases, the virus was fatal in itself. Multiplying explosively throughout the respiratory tract, it provoked an immune response so furious that it devastated the lung's delicate tissues. And it was those deaths that explained H1N1's unique terror. Influenza typically kills the very young and the old, whose immune systems are too weak to fight it off, but Spanish flu killed young men and women in the prime of life.

The exceptional virulence of Spanish flu is not fully explained, although Jeffery Taubenberger, a military pathologist, has been reconstructing the viral genome from Army autopsy samples and a corpse preserved in the Alaskan permafrost. (His preliminary finding is that it represented only minor changes from a variety known to infect birds.) But researchers understand this much: the H1 subtype was new to most humans then alive, who as a result were almost defenseless against it. The same was true of the H2 subtype in 1957, and of H3 in 1968. "Whenever a virus shows up in humans with a new [H factor], you get a pandemic," says Treanor. That is one reason why doctors are so wary of an H5 virus: hardly anyone alive today has been exposed to one. "We thought we knew the rules," says Stephen Morse, director of the Center for Public Health Preparedness at Columbia University, "and one of those rules was that H1, H2 and H3 cause flu in humans, not H5. This is like the clock striking 13."

There is, fortunately, one big barrier that H5N1 apparently hasn't crossed yet: efficient human-to-human transmission. No one knows what molecular rearrangement might permit it to crash this barrier, but it could happen in a couple of ways: by random mutation, or by exchanging genetic information with a mammalian virus, either in a human host or another animal, such as a pig. Many flu variants are believed to have emerged in that way, and there are plenty of opportunities. "The ominous aspect of avian flu is how widespread the virus is among poultry in Eastern Asia," says Dr. Julie Gerberding, director of the Centers for Disease Control and Prevention. "From a statistical perspective, we are looking at a very worrisome situation." A 1999 study by the CDC forecast that a new flu pandemic of "average" lethality could cause as many as 207,000 deaths in this country alone, with up to 100 million people infected. Direct medical costs could reach $166 billion.

Of course, an H5N1 pandemic could be much worse, or it could be mild. Some researchers, like Paul Ewald, a biologist at the University of Louisville, doubt that another 1918 is at hand, even if the virus mutates. There were unique circumstances that year; the war forced millions of young men into close quarters in barracks, troopships and trenches, where there was no escape from a sick comrade. In normal circumstances, Ewald speculates, the spread of Spanish flu would have been limited by its very virulence. People too sick to get up and move around--not to mention those who die--don't infect others, unless they're sleeping inches apart. In the absence of war, he presumes the flu would have either burned itself out quickly or evolved to a less virulent form that could spread more efficiently.

Still, that's just a hypothesis, and not one to which you'd choose to entrust your grandmother's life. So researchers are trying to decode the H5N1 genome, looking for clues to its virulence and its possible vulnerabilities. And while they do, trillions of blindly replicating viruses are mutating and combining in unpredictable ways. In fact, as Richard Webby of St. Jude Children's Research Hospital in Memphis, Tenn., points out, of the eight gene segments in the virus, only one--the H5 gene itself--remains from the version that first began killing people in 1997. "It's a completely different beast," he says, although whether it's becoming more deadly, he cannot say. But just last week, 30 captive Bengal tigers on a farm in Thailand died after being fed raw chickens contaminated with avian flu. Thoroughly cooking poultry kills the virus, but the keepers weren't aware the chickens were infected. Or perhaps they just weren't concerned, because flu viruses were almost never deadly to cats. Until this one.